Slow first "copy job" with WAN accelerator

[crond]

Hi All,

I added WAN accelerators to backup-copy job and the first run (after adding WAN accelerators) is extremely slow

Source WAN is running on the VBR11 server (6cores/16G RAM / Win server 2019 VM)
and according to VBR console is main bottleneck. The KB/help page recomends to allocate at least 8GB to WAN accelerator, so I thought 16G should be enough but apparently this is not the case. Looking at resource monitor WAN accelerator consumes 95% of system RAM and swaps like crazy.
So how much RAM does it actually need ?

[crond]

A follow-up question ... How is "bottleneck" calculated ?
The job reports
Source / Source WAN / Network / Target WAN / Target = 0% / 99% / 1% / 99% / 0%

While I see a lot of activities on Source WAN I don't see any activity on Target WAN .. CPU ~1% / RAM ~under 20% / no disk activity .. so how is it 99% ?

[Mildur]

Hi Dmitry

Our FAQ has more information about the bottleneck analysis.

With 99% on both accelerators (source and target), it seems to me that the bottleneck lies between processing the data on the source WAN accelerator and speaking to the target WAN accelerator. What's your bandwidth and in which mode (low or high) does the WAN accelerator operate?
If you use the global cache with the low-bandwidth mode, the first run can take longer because Veeam has to populate the global cache.

I think it's best to let support analyze the debug logs, if you don't see any progress as expected. If you open a support case, please let us know the case number.

Thanks
Fabian

[crond]

Hi Fabian,

Thanks for reply. May I ask you to also answer the first question. How much RAM does source WAN need ? I looked through FAQ and also the "WAN accelerator sizing" KB, but there's nothing about required RAM size.

To answer your questions the WAN is in "low" bandwidth mode.
VM size is 10TB and mostly static, with under 100GB/week worth of changes. The backup bandwidth is limited to 10Mbps (thus the desire to use WAN accelerator) and initial seeding was done manually. I.E initial backup copy job was run locally, backup files were transferred through other channel to target repository. After mapping backup at the final destination back to the job, wan accelerators were added to backup-copy job.